The success of pervasive business intelligence is all about trust in data. Enterprise data warehousing plays a crucial role in delivering information so that users can drive real-time decision-making reliably, and accurately reflects existing business conditions.
However, the enterprise information needed for pervasive BI comes in many forms, exists in many silos, and has many degrees of latency. The challenge is, before we hand off decision-making to more automated processes, how can we be sure this information presents a true picture of business? You need a stronger foundation to ensure freshness, completeness and accuracy of data.
These are the issues that James Markarian, CTO of Informatica, along with Stephen Brobst, CTO of Teradata, pondered and explored in a recent Webcast.
The goal of pervasive BI is to serve up just-in-time insights when and where they are needed. It may be based on historical data coming from the data warehouse, it may be based on real-time data coming right in from a production system, or it could likely be a combination of both. Pervasive BI promises to blend analytical data with real-time transactional data, analyze it on the fly, and deliver the information to front-line employees and managers. Information latency directly impacts business results.
Moving to pervasive BI may seem like a daunting task, but the good news is that it all doesn’t have to be done at once. Markarian said that moving to pervasive BI need not be an all-encompassing project affecting all data across the enterprise. “Pervasive BI doesn’t mean an immediate wholesale change to the way that we do everything,” he explained. “It’s an architecture and framework.” Most organizations take an incremental approach starting with departmental data warehousing. They can then expand their practice to involve enterprise-wide deployment
However, many of today’s enterprise data management solutions and platforms are configured to operate within silos serving specific processes or functions. “Enterprise application integration, enterprise service buses, and complex event processing technologies – when you had siloed decision making, this was all fine,” Markarian says. “In pervasive business intelligence, you’re looking for something a lot more holistic, where you’re combining information with different latencies. This is the real-time index with the pervasive view. We need to create a single architecture that can handle all workloads.”
The way to do this is through building out an integration or data services layer that will support and rationalize data coming in from varied sources with various degrees of latency. Such a service layer needs to be able to ultimately support terabytes’ worth of information, Markarian says.
Pervasive BI is also helping to make the long-awaited “democratization” of business intelligence a reality – moving capabilities beyond specialized business analysts workstations and out to all levels of employees. However, this also calls for a new breed of BI tools that perform analysis in tandem with transactions processed by applications. Brobst observes that the traditional BI tools used within enterprises are built for “exploratory strategic analysis,” not for pervasive BI. The proliferation of pervasive BI is “not going to happen by deploying traditional BI tools to the people on the front line,” he says.
Instead, Brobst sees more BI functionality delivered through “decisioning services” made available to the entire enterprise, both inside and outside the firewall. Brobst observes that data warehouses – active data warehouses – will be an important, but just one part, of this new equation. “In my mind, the most important thing about the reference architecture is that in the middle of this picture is not a ‘fat’ data warehouse,” he says. “The success of delivering pervasive BI really depends on our ability to get the data warehouse into the mainstream of our corporate IT infrastructure, and plug into business services automation capabilities, so we can leverage those capabilities to a much wider audience within our enterprises.”
Brobst emphasized that pervasive BI should not be confused with a form of “real time data warehousing,” noting that “people get trapped into the mindset of ‘What I’m going to do is get the data faster.'” The true goal of pervasive BI should be “to reduce the time between business events, and our ability to take informed action based on those business events,” he says.
For years, we’ve heard the promise of “BI for the masses,” but this has not yet come to pass. In surveys I have seen and conducted myself, most BI activity has remained confined to specialists, working with historical data to look at things that have just happened and extrapolating what may soon happen within the business. Since pervasive BI links analytical capabilities to front-line applications, and can be delivered through a service oriented architecture, this may finally mean that BI is advancing in a meaningful way into the enterprise. What is your strategy for ensuring your data readiness for Pervasive BI?